I have three versions of the functions to do element-wise comparison of two lists and output a count of results. First uses for loop (simple function), second uses list expression, third uses numpy. I expected numpy to be super-fast especially once when the list sizes are large, but find its not fast consistently.
Running this on google colab with different array sizes gives me results as follows: SN) ArraySize RuntimeRatio=Simple:Optimised:Numpy
- 25 1:0.43:2.61
- 25 1:0.22:0.46
- 25 1:0.63:0.29
- 25 1:0.75:1.18
- 2500 1:0.89:3.07
- 2500 1:0.84:1.51
- 2500 1:0.59:0.79
- 2500 1:0.75:2.19
- 250000 1:1.26:2.64
- 250000 1:1.23:2.18
- 250000 1:1.25:2.22
- 250000 1:0.90:1.56
- 25000000 1:1.40:2.25
- 25000000 1:1.32:2.22
- 25000000 1:1.29:2.17
- 25000000 1:1.28:2.19
Any ideas on whats happening or what I am doing wrong?
The code:
import numpy as np
import time
import random
def solution_list_simple(a, b):
answer = 0
for aval, bval in zip(a, b):
if aval > bval:
answer += 1
return answer
###########
def solution_list_opti(a, b):
return [ a_ele > b_ele for a_ele, b_ele in zip(a, b) ].count(True)
###########
def solution_np(a, b):
return np.sum( np.array(a) > np.array(b) )
###########
random.seed(30)
howmanyvalue = 250000
A = [ random.randint(1, 100) for _ in range(howmanyvalue) ]
B = [ random.randint(1, 100) for _ in range(howmanyvalue) ]
## list version - simple
start_time = time.time()
print(f"")
print(f"\nanswer list simple = {solution_list_simple(A, B)}")
runtime_list_simple = time.time() - start_time
print(f"Runtime list simple = {runtime_list_simple}")
## list version - optmised
start_time = time.time()
print(f"")
print(f"\nanswer list_opti = {solution_list_opti(A, B)}")
runtime_list_opti = time.time() - start_time
print(f"Runtime list optimised = {runtime_list_opti}")
## numpy version
start_time = time.time()
print(f"")
print(f"\nanswer numpy = {solution_np(A, B)}")
end_np = time.time()
runtime_numpy = time.time() - start_time
print(f"Runtime numpy = {runtime_numpy}")
print(f"\n\nRelative ratios\nlist_simple : list_optimised : numpy = {1} : {(runtime_list_opti / runtime_list_simple):.2f} : {(runtime_numpy / runtime_list_simple):.2f}")
